dma and demo kernels

kernel/include/gemmini_mmio.h

@@ -5,7 +5,8 @@
 #endif
 
 #define SMEM_BASE 0xff000000
-#define SMEM_SIZE 0x4000
+//#define SMEM_SIZE 0x4000
+#define SMEM_SIZE 0x10000
 
 #define SMEM_MASK (SMEM_SIZE - 1)
 #define SMEM_ADDR_END (SMEM_BASE + SMEM_SIZE)
@@ -31,10 +32,13 @@
 // 0: k = 0, no accumulation
 // 1: k % 2 = 0, buffer regions 0
 // 2: k % 2 = 1, buffer regions 1
-// 8, 9: memory ops
-// 8: tile-sized move-in (unused)
+// 8, 9, 10, 11: memory ops
+// 8: tile-sized move-in stride
 // 9: tile-sized move-out
+// 10: tile-sized move-in, buffer regions 0
+// 11: tile-sized move-in, buffer regions 1
 #define GEMMINI_CISC_CMD_I(x) asm("csrwi 0xacc, "#x)
+#define GEMMINI_CISC_CMD_R(x) asm("csrw 0xacc, %0" :: "r" (x))
 #define GEMMINI_STATUS() ({uint32_t status; asm volatile ("csrr %0, 0xacc" : "=r" (status)); status;})
 
 // convert normal matrix i,j into tiled smem offset
@@ -62,6 +66,9 @@
     /* sprintf((char *) PRINT_BUF, "%llx %llx %d\n", rs1, rs2, funct); */ \
 }
 
+#define loop_matmul_skips(skip_lda, skip_ldb, skip_ldd, skip_ex, skip_stc) \
+  (((skip_lda) | ((skip_ldb) << 1) | ((skip_ldd) << 2) | ((skip_ex) << 3) | ((skip_stc) << 4)) << 3)
+
 #define sp_tiled_matmul_full_spad_ws(A_sp_addr_start, B_sp_addr_start, D_sp_addr_start, C_dst_sp_addr_start,\
   I, J, K, pad_I, pad_J, pad_K, a_transpose, b_transpose, full_C, low_D, acc, act, skips) \
   gemmini_loop_ws_spad(I, J, K, pad_I, pad_J, pad_K, A_sp_addr_start, (B_sp_addr_start) + (K) * (J) * DIM, NULL, \

kernel/linker/vx_link32.ld

@@ -10,8 +10,8 @@ ENTRY(_start)
 
 MEMORY {
   DRAM0    (rwx): ORIGIN = 0x80000000, LENGTH = 512M
-  DRAM1    (rwx): ORIGIN = 0xa0000000, LENGTH = 32K
-  DRAM2    (rwx): ORIGIN = 0xa1000000, LENGTH = 32K
+  DRAM1    (rwx): ORIGIN = 0xa0000000, LENGTH = 16M
+  DRAM2    (rwx): ORIGIN = 0xa1000000, LENGTH = 16M
 }
 
 SECTIONS
@@ -69,6 +69,7 @@ SECTIONS
     /* .gnu.warning sections are handled specially by elf.em.  */
     *(.gnu.warning)
   }
+
   .fini           :
   {
     KEEP (*(SORT_NONE(.fini)))
@@ -128,6 +129,7 @@ SECTIONS
     KEEP (*(.init_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .ctors))
     PROVIDE_HIDDEN (__init_array_end = .);
   }
+
   .fini_array    :
   {
     PROVIDE_HIDDEN (__fini_array_start = .);
@@ -135,6 +137,18 @@ SECTIONS
     KEEP (*(.fini_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .dtors))
     PROVIDE_HIDDEN (__fini_array_end = .);
   }
+
+  .htif_pad : {
+    . = ALIGN(0x1000);
+  }
+
+  . = ALIGN(0x1000);
+  .tohost : {
+    *(.tohost)
+    /* . += 0x100; */
+  }
+
+  . = ALIGN(0x1000);
   .ctors          :
   {
     /* gcc uses crtbegin.o to find the start of
@@ -216,6 +230,8 @@ SECTIONS
 		            MAX(__DATA_BEGIN__ + 0x800, __BSS_END__ - 0x800));
   _end = .; PROVIDE (end = .);
   . = DATA_SEGMENT_END (.);
+
+
   /* Stabs debugging sections.  */
   .stab          0 : { *(.stab) }
   .stabstr       0 : { *(.stabstr) }

kernel/src/vx_start.S

@@ -150,3 +150,4 @@ vx_wspawn_wait:
 	.weak __dso_handle
 __dso_handle:
 	.long	0
+

kernel/tohost.S

@@ -0,0 +1,7 @@
+.section ".tohost","aw",@progbits
+.align 6
+.globl tohost
+tohost: .dword 0
+.align 6
+.globl fromhost
+fromhost: .dword 0